An Embedded Edge Architecture for Real-Time Acoustic Direction Monitoring Using a Four-Microphone Array and GCC-PHAT

This paper presents SoundSL, a low-cost embedded acoustic direction-of-arrival (DoA) indicator based on an ESP32 and a four-microphone digital MEMS array acquired via dual I2S. The system estimates the azimuth of a sound source in real time and drives an eight-sector LED ring for immediate user feedback. A computationally efficient hybrid pipeline is proposed: the microphone pair with the highest signal power is selected, a GCC-PHAT (generalized cross-correlation with phase transform) time-difference-of-arrival (TDOA) estimate is computed on the chosen pair, and pair-dependent calibration plus stabilization logic (adaptive thresholds, IIR smoothing, two-hit confirmation, and silence hold) are applied before mapping the result to the LED sectors. The prototype was evaluated by rotating the device in 10° steps and collecting 30 independent decisions per angle using a compact loudspeaker at 32.5 cm. Across 36 tested angles, the system achieved a mean correct-detection rate of 80.1%, with a minimum of 53.33% and a maximum of 100%. The results show that reliable DoA indication can be obtained on a resource-constrained microcontroller using phase-based correlation coupled with lightweight robustness logic, enabling practical assistance and IoT sensing applications.